Graph based RecuroMatch Segmentation Algorithm in Citrus Fruit Images

London Journal of Research in Computer Science and Technology
Volume | Issue | Compilation
Authored by Renuga Devi , V. Kavitha
Classification: D.2.6
Keywords: citrus, graph segmentation, recuromatch, features, kernel points.
Language: English

Graph segmentation is promising as a very effective process for learning the complex structures and relationships hidden in disease data. Segmentation of citrus fruit diseases is a major task of image processing. There is no common segmentation process that can deal with all diseased portions, and the correct solution will always to a certain degree depend on subjectivity. To solve these issues, in this paper to develop a novel Graph based RecuroMatch (GRM) segmentation algorithm to discover citrus fruit diseases with different illumination conditions. To identify the diseased regions in citrus fruit, the GRM algorithm has to be described. The proposed work of a citrus fruit segmentation process presents three tasks namely, i) Image pre-processing: it is carried out using remove the irrelevant noises; ii) Citrus fruit features extraction: Feature extraction using new Colpromatix color space model, Size, Texture, Shape, and Coarseness; and iii) Graph based RecuroMatch segmentation process is an important process to discover the disease feature of an image. Segmentation process is playing an important function in systems for disease portion recognition, extracting, and examination.

               

Graph based RecuroMatch Segmentation Algorithm in Citrus Fruit Images

V. Kavithaα  &  Dr. M. Renuka Deviσ

____________________________________________

  1. ABSTRACT 

Graph segmentation is promising as a very effective process for learning the complex structures and relationships hidden in disease data. Segmentation of citrus fruit diseases is a major task of image processing. There is no common segmentation process that can deal with all diseased portions, and the correct solution will always to a certain degree depend on subjectivity. To solve these issues, in this paper to develop a novel Graph based RecuroMatch (GRM) segmentation algorithm to discover citrus fruit diseases with different illumination conditions. To identify the diseased regions in citrus fruit, the GRM algorithm has to be described. The proposed work of a citrus fruit segmentation process presents three tasks namely, i) Image pre-processing: it is carried out using remove the irrelevant noises; ii) Citrus fruit features extraction: Feature extraction using new Colpromatix color space model, Size, Texture, Shape, and Coarseness; and iii) Graph based RecuroMatch segmentation process is an important process to discover the disease feature of an image. Segmentation process is playing an important function in systems for disease portion recognition, extracting, and examination.

Keywords: citrus, graph segmentation, recuro- match, features, kernel points.

Author α: P.hd Scholar, Bharathiar University Coimbatore, India.

σ: Director of MCA, Sri Venkateswara College of Computer Applications and Management, Ettimadai Coimbatore, India.

  1. INTRODUCTION

Image Segmentation partitions an image into distinct regions containing each pixel with similar attributes. To be meaningful and useful for image analysis and interpretation, the regions should strongly relate to depicted objects or features of interest [13]. Meaningful segmentation is the first step from low-level image processing transforming a grayscale or color image into one or more other images to high-level image description in terms of features, objects, and scenes. The success of image analysis depends on reliability of segmentation, but an accurate partitioning of an image is generally a very challenging problem.

The goal of citrus image segmentation is to cluster pixels into significant image regions, i.e., regions corresponding to individual surfaces, objects, or natural parts of objects. The objective of citrus fruit segmentation is to detect and extract the regions which compose an image. Note that different to the classification problem [14], recognition [11] of these regions is not required. Hence, it is not possible to identify the different objects, simply because it is the initial time. So, the solution of the citrus fruit segmentation process will be something as: “there are four regions (Background, foreground, particles and orientations) in the image” and an array of the size of the image where each pixel is labeled with the corresponding region number.

The most essential segmentation process that may be carried out is thresholding of an image (see [1]). This method consists on comparing the measure associated to each pixel to one or some thresholds in order to determine the class which the pixel belongs to. The attribute is generally the grey level, although color or a simple texture descriptor can also be used. Thresholds may be applied globally across the image (static threshold) or may be applied locally so that the threshold varies dynamically across the image [10].

Efficient graph-based image segmentation, introduced in [2] by Felzenszwalb and Huttenlocher, is an additional method of performing clustering in feature space. This method works directly on the data points in feature space, without first performing a filtering step, and uses a variation on single linkage clustering. The key to the success of this method is adaptive thresholding. To perform traditional single linkage clustering, a minimum spanning tree of the data points is first generated (using Kruskal’s algorithm), from which any edges with length greater than a given hard threshold are removed. The connected components become the clusters in the segmentation. The method in [2] eliminates the need for a hard threshold, instead replacing it with a data-dependent term.

For a segmentation algorithm to be a useful ‘graph based RecuroMatch (GRM)’ in a larger system, we propose that it should have three crucial characteristics:

  • Correctness: the ability to generate segmentations which agree with ground truth. That is, neither segmentations which correctly recognize configuration in the image at neither too fine nor too coarse a level of detail.
  • Stability with respect to parameter choice: the ability to create segmentations of reliable exactness for a collection of parameter choices.
  • Stability with respect to image choice: the ability to create segmentations of reliable correctness using the similar parameter selection on special images.

If a segmentation method satisfies these three requirements, then it will provide positive and conventional results which can be consistently included into a larger system exclusive of excessive parameter tuning. It has been argued that the correctness of a segmentation algorithm is only relevant when measured in the context of the larger system into which it will be incorporated. However, most such systems assume that a segmentation algorithm satisfies a subset of the criteria above. In addition, there is value in preparing out algorithms which give nonsensical results and limiting the list of possible algorithms to those that are well-behaved even if the components of the rest of the system are unknown.

The aim of this work is to develop a citrus fruit segmentation process is to efficiently clustering objects in images to assist fruit deficiency detection. In this paper, we present citrus fruit image segmentation scheme using graph based RecuroMatch technique is used to segment the feature extracted image because it is known to provide a good quality segmentation result for further classification process.

The rest of the paper is organized as follows: Literature Review is detailed in Sect. 2. In Sect. 3, Research methodologies acquire citrus fruit image segmentation process and conclusion is in Sect. 4.

  1. LITERATURE REVIEW

A transform for multi-scale image segmentation by integrated edge and region detection [3] proposed transform avoids this by letting the structure emerge, bottom-up, from interactions among pixels. The transform involves global computations on pairs of pixels followed by vector integration of the results. An attraction force field is computed over the image in which pixels belonging to the same region are mutually attracted and the region is characterized by a convergent flow. It is shown that the transform possesses properties that allow multi-scale segmentation, or extraction of original, un-blurred structure at all different geometric and photometric scales present. This is in contrast with much previous work wherein multi-scale structure is viewed as the smoothed structure in multi-scale signal decimation. Scale is an integral parameter of the force computation, and the number and values of scale parameters associated with the image can be estimated automatically. Regions are detected at all a priori unknown scales resulting in automatic construction of a segmentation tree, in which each pixel is annotated with descriptions of all the regions it belongs to. Transform properties are presented for piecewise-constant images but hold for more general ones

A two-stage process for accurate image segmentation [3] illustrated segmenting images is one of the most important steps in many high-level computer vision algorithms. The ability to divide images up into meaningful regions based upon properties such as shape, texture and color has still not been fully solved. In this paper authors showed how good quality segmentations of complex outdoor scenes may be achieved by a two-stage process. The first step is to produce an approximate segmentation using only color and texture information. The second step is to merge regions by using a neural network trained to classify the regions into one of eleven possible types which correspond to objects types found in outdoor scenes. This stage involves the use of high-level knowledge such as position, shape, context and orientation.

Image segmentation using expectation- maximization and its application to image querying [4] has described a system that uses the Blobworld representation to retrieve images from this collection. An important aspect of the system is that the user is allowed to view the internal representation of the submitted image and the query results. Similar systems do not offer the user this view into the workings of the system; consequently, query results from these systems can be inexplicable, despite the availability of knobs for adjusting the similarity metrics. By finding image regions that roughly correspond to objects, they allow querying at the level of objects rather than global image properties. They presented results indicating that querying for images using Blobworld produces higher precision than does querying using color and texture histograms of the entire image in cases where the image contains distinctive objects.

Discriminative re-ranking of diverse segmentations [6] introduced a hybrid, two-stage approach to semantic image segmentation. In the first stage a probabilistic model generates a set of diverse plausible segmentations. In the second stage, a discriminatively trained re-ranking model selects the best segmentation from this set. The re-ranking stage can use much more complex features than what could be tractably used in the probabilistic model, allowing a better exploration of the solution space than possible by simply producing the most probable solution from the probabilistic model. While the proposed approach already achieves state-of-the-art results (48%) on the challenging VOC 2012 dataset, machine and human analyses suggest that even larger gains are possible with such an approach.

Holistically-nested edge detection [7] developed a new edge detection algorithm that tackles two important issues in this long-standing vision problem: (1) holistic image training and prediction; and (2) multi-scale and multi-level feature learning [13]. The authors proposed, holistically-nested edge detection (HED), performs image-to-image prediction by means of a deep learning model that leverages fully convolution neural networks and deeply- supervised nets. HED automatically learns rich hierarchical representations (guided by deep supervision on side responses) that are important in order to approach the human ability resolve the challenging ambiguity in edge and object boundary detection. To significantly advance the state-of-the-art on the BSD500 dataset (ODS F-score of .782) and the NYU Depth dataset (ODS F-score of .746), and do so with an improved speed (0.4 second per image) that is orders of magnitude faster than some recent CNN-based edge detection algorithms.

Region-based convolution networks for accurate object detection and semantic segmentation [8] proposed a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 50 percent relative to the previous best result on VOC 2012-achieving a mAP of 62.4 percent. Our approach combines two ideas: (1) one can apply high-capacity convolutional networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data are scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, boosts performance significantly. Since to combine region proposals with CNNs, call the resulting model an R-CNN or Region-based Convolutional Network.

Multi-view learning with incomplete views [9] suggested that the key to handling the incomplete-view problem is to exploit the connections between multiple views, enabling the incomplete views to be restored with the help of the complete views. They proposed an effective algorithm to accomplish multi-view learning with incomplete views by assuming that different views are generated from a shared subspace. To handle the large-scale problem and obtain fast convergence, to investigate a successive over-relaxation method to solve the objective function. Convergence of the optimization technique is theoretically analyzed. The experimental results on toy data and real-world data sets suggest that studying the incomplete-view problem in multi-view learning is significant and that the proposed algorithm can effectively handle the incomplete views in different applications.

  1. RESEARCH METHODOLOGY

The research methodology considers the Citrus fruit segmentation process of graph based RecuroMatch algorithm is derived in this section. We introduce a tractable segmentation process of RecuroMatch algorithm from feature extracted (Size, Shape, Texture and Coarseness) images, which is a natural extension of image feature extraction process. The overall feature extraction process flow diagram is described in figure 1.

Fig. 1: Proposed Flow Diagram

4.1  Image Preprocessing

Image preprocessing is a mining technique that performs transforming raw image data into a reasonable format. In this process, original images pixels size (1027 x 768 x 3) is resized into (256 x 256 x 3) dimensions without pixel loss using ‘bicubic’ method. After that, images must be of the same size and are supposed to be associated with indexed images on a common color map.

4.2  Image Feature Extraction

Image feature extraction is done without local decision making; the result is often referred to as a feature image.  Consequently, a feature image can be seen as an image in the sense that it is a function of the same spatial (or temporal) variables as the original image, but where the pixel values hold information about image features instead of intensity or color. Feature extraction is a mining technique that involves transforming raw data into a comprehensible format. Feature extraction is a proven method of resolving Colpromatix Color code, Texture Shape, and Coarseness.

The Color feature extraction result finds the disease portions 90% located clearly in black color represents in figure 2.

Fig. 2: Result of Colpromatix Color Feature Extraction; First column shows input RGG Color space mode; Second column shows Gray Color Space; Third column shows the Colpromatix Color space result model

4.3  Graph based RecuroMatch Algoirthm

Graph based RecuroMatch algorithm mostly rely on the assumption that the neighboring pointer pixels within multiple region have similar value. The common process is to compare single pixel with its neighbors. If a similarity criterion is satisfied, the pixel can be set belong to the cluster as one or more of its neighbors.

A RecuroMatch algorithm is an extension of region-based method usually proceeds as follows:

  • The image is partitioned into connected regions by grouping neighboring pixels of similar intensity levels.
  • Adjacent regions are then merged under some criterion such as homogeneity or sharpness of region boundaries.
  • Over stringent criteria create fragmentation; lenient ones overlook blurred boundaries and over merge.
  • The proposed RecuroMatch is locate the disease boundaries based on the Region Pointers (RP*) according to the overlook boundaries

The Graph based RecuroMatch (GRM) algorithm is one of the novel extension of region-based segmentation. It performs a segmentation of an image with observe the neighboring pixels of a group of points, known as kernel (seed) points, and determine whether the pixels could be classified to the cluster of seed point or not [3]. The algorithm procedure is as follows,

Algorithm 1: Graph RecuroMatch Process

Step 1: To start with a amount of kernel points which have been clustered into m clusters, called C1, C2, …, Cm. And the locations of initial kernel points is set as p1, p2, …, p3.

Step 2: To compute the difference of pixel value of the initial kernel point pi and its neighboring points, if the difference is lesser than the feature value (colpromatix feature) we define, the neighboring point could be classified into Ck , where k = 1, 2, …,m.

Step 3:

if (match == feature_right_dimension)

      if (!=feature_boundary)

            {return false;}

      feature_boundary++;

  Match++;         

Step 4:

if (match == feature_left_dimension)

    if (RecuroMatch (feature_right_dimension, Match + 1))

     return true;

 if (RP*feature_right_dimension && RecuroMatch (feature_left_dimension + 1, Match))

       return true;

  else      

       return false;

Step 5: Re-compute the boundary of Ck and set those boundary points as new kernel points pi(k). In addition, the mean pixel values of Ck have to be recomputed, respectively.

Step 6: Repeat Step 2 and 4 until all pixels in image have been allocated to a suitable group.

The advantage of proposed GRM algorithm is no predefined threshold is made by user and it usually based on color feature (colpromatix) intensity values. The segmented regions are chosen to be as consistent as possible. There is no doubt that each of the segmentation regions of GRM has high similarity and no disconnected problem. The proposed GRM results are described in Figure 3 to 6.

Fig. 3: Input Citrus Diseased Fruit Image-1

Fig. 4: Graph based RecuroMatch Segmentation Result

Fig. 5: Input Citrus Diseased Fruit Image-2

Fig. 6: Graph based RecuroMatch Segmentation Result

V.    CONCLUSION

In this paper reviewed and executed the progression of the information and communication knowledge in the field of citrus image disease segmentation. Citrus fruit image processing approaches used in the field of agriculture and food industry for fruit classification of two processes is explored in this paper. Most of the work in this citrus image segmentation processing is composed of the mainly three main steps (1) Image preprocessing (2) feature extraction and (3) Graph RecuroMatch segmentation. In the first and second steps image preprocessing and citrus feature extraction process is carried out using Colpromatix method. In this paper mainly focused on novel segmentation algorithm of Graph RecuroMatch (GRM) process are segment the disease portions and figured out. The proposed GRM method uses special key pointers features in graph segmentation to find out the disease locations effectively.

The further work is to do post-processing of data mining algorithms to classify diseases based advanced classification algorithms.

REFERENCES

  1. P.K. Sahoo, S. Soltani, A.K.C.Wong, and Y.C. Chen. A survey of thresholding techniques. Computer Vision, Graphics and Image Processing, 41: 233–260, 1988.
  2. P. Felzenszwalb and D. Huttenlocher. Efficient graph-based image segmentation. IJCV, 59(2): 167–181, 2004.
  3. N. Ahuja. A transform for multiscale image segmentation by integrated edge and region detection. PAMI, 1996.
  4. N.W. Campbell, B. T. Thomas, and T. Troscianko. A two-stage process for accurate image segmentation. In Int. Conf. on Image Processing and its Applications, 1997.
  5. C. Carson, S. Belongie, H. Greenspan, and J. Malik. Blobworld: Image segmentation using expectation-maximization and its application to image querying. PAMI, August 2002.
  6. P. Yadollahpour, D. Batra, and G. Shakhnarovich, “Discriminative re-ranking of diverse segmentations,” in Proc. IEEE Conf. CVPR, Jun. 2013, pp. 2099–2104.
  7. S. Xie and Z. Tu, “Holistically-nested edge detection,” in Proc. ICCV, 2015, pp. 1395–1403.
  8. R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Region-based convolutional networks for accurate object detection and semantic segmentation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 38, no. 1, pp. 142–158, Jan. 2016.
  9. C. Xu, D. Tao, and C. Xu, “Multi-view learning with incomplete views,” IEEE Trans. Image Process., vol. 24, no. 12, pp. 5812–5825, Dec. 2015.
  10. P. Felzenszwalb and D. Huttenlocher. Efficient graph-based image segmentation. IJCV, 59(2): 167–181, 2004.
  11. P. Felzenszwalb and D. Huttenlocher. Pictorial structures for object recognition. IJCV, 61, 2005.
  12. R. Fergus, P. Perona, and A. Zisserman. Object class recognition by unsupervised scale- invariant learning. In CVPR, volume 2, pages 264–271, June 2003.
  13. C. Fowlkes, D. Martin, and J. Malik. Learning affinity functions for image segmentation: Combining patch-based and gradient-based approaches. In CVPR, 2003.
  14. K. Grauman and T. Darrell. The pyramid match kernel: Discriminative classification with sets of image features. In ICCV, 2005.



author

For Authors

Author Membership provide access to scientific innovation, next generation tools, access to conferences/seminars
/symposiums/webinars, networking opportunities, and privileged benefits.
Authors may submit research manuscript or paper without being an existing member of LJP. Once a non-member author submits a research paper he/she becomes a part of "Provisional Author Membership".

Know more

institutes

For Institutions

Society flourish when two institutions come together." Organizations, research institutes, and universities can join LJP Subscription membership or privileged "Fellow Membership" membership facilitating researchers to publish their work with us, become peer reviewers and join us on Advisory Board.

Know more

subsribe

For Subscribers

Subscribe to distinguished STM (scientific, technical, and medical) publisher. Subscription membership is available for individuals universities and institutions (print & online). Subscribers can access journals from our libraries, published in different formats like Printed Hardcopy, Interactive PDFs, EPUBs, eBooks, indexable documents and the author managed dynamic live web page articles, LaTeX, PDFs etc.

Know more