A disease spot segmentation method using comprehensive color feature with multi-resolution channel and region growing
- PDF / 1,623,683 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 84 Downloads / 175 Views
A disease spot segmentation method using comprehensive color feature with multi-resolution channel and region growing N. Jothiaruna 1 & K. Joseph Abraham Sundar 1
& M. Ifjaz Ahmed
1
Received: 11 June 2020 / Revised: 7 August 2020 / Accepted: 15 September 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
The paper proposes a novel method to segment disease spots in leaves using image processing techniques. In the process of disease spot segmentation many challenges are faced such as uneven illumination and clutter background. To solve uneven illumination, color spaces and gray scale conversions are summed. Color spaces like H (hue) component of HSV, L*a*b* color spaces and Excess Red index (ExR) are used. Color to gray scale conversion is done by using weighted mulitresolution channel. Region growing method is used to solve the clutter background issues by interactively selecting growing seeds under real field conditions. Using precision, performance measure is calculated and an average segmentation accuracy of 94% is achieved. Keywords Segmentation . Color spaces . Grayscale conversion . Region growing method . Disease spots
1 Introduction In agriculture field, disease detection in plant plays an important role. The preliminary method of disease leaf detection is subtracting low pass filtered image from the original image to remove pixel to pixel correlation. Both laplacian and Gaussian filter are used, but laplacian pyramid requires simple computation for several levels [2]. In [5] Sampling an image, reducing a dimension and color difference are tested by Gaussian pairing, this techniques are useful very much in real time process. It has many properties like predicting a hue, luminance, saturation. Optimizing non-linear global mapping to get original color brightness i.e., it robustly produces the visual appearance of RGB image in grayscale image [10]. Fusion of enhanced image process is carried out here [3], pattern
* K. Joseph Abraham Sundar [email protected]
1
Computer Vision and Soft Computing Lab, School of Computing, SASTRA Deemed University, Thanjavur, Tamil Nadu, India
Multimedia Tools and Applications
match is used to control the image combination and averaging, then match measuring for past implementation is defined as neighborhood of every pyramid instead of pyramid itself. Using the filter theory channel level distinction is measured to frame channel salience to represent the filter level of RGB color [23]. Based on mulitresolution contrast merging an image scheme is introduced and this scheme is tested by using merging the thermal and visual image. Tested result shows that fused image has detailed representation of described scene [19]. A two dimensional color histogram and support vector machine (SVM) method for pixel wise classification and quantification of disease is proposed in [22]. Converting color image to grayscale image is done using Singular Value Decomposition (SVD) by incorporating chrominance information from RGB image i.e., converting thre
Data Loading...
