SVM-PSO based rotation-invariant image texture classification in SVD and DWT domains

2016 ◽  
Vol 52 ◽  
pp. 96-107 ◽  
Author(s):  
Bae-Muu Chang ◽  
Hung-Hsu Tsai ◽  
Chih-Yuan Yen
Keyword(s):  
Texture Classification ◽  
Image Texture ◽  
Rotation Invariant

CNN performance dependence on linear image processing

2020 ◽  
Vol 2020 (10) ◽  
pp. 310-1-310-7
Author(s):  
Khalid Omer ◽  
Luca Caucci ◽  
Meredith Kupinski
Keyword(s):  
Image Processing ◽  
Texture Classification ◽  
Full Rank ◽  
Detection Performance ◽  
Ideal Observer ◽  
Training Data ◽  
Image Texture ◽  
Training Images ◽  
Analytic Expressions ◽  
Linear Compression

This work reports on convolutional neural network (CNN) performance on an image texture classification task as a function of linear image processing and number of training images. Detection performance of single and multi-layer CNNs (sCNN/mCNN) are compared to optimal observers. Performance is quantified by the area under the receiver operating characteristic (ROC) curve, also known as the AUC. For perfect detection AUC = 1.0 and AUC = 0.5 for guessing. The Ideal Observer (IO) maximizes AUC but is prohibitive in practice because it depends on high-dimensional image likelihoods. The IO performance is invariant to any fullrank, invertible linear image processing. This work demonstrates the existence of full-rank, invertible linear transforms that can degrade both sCNN and mCNN even in the limit of large quantities of training data. A subsequent invertible linear transform changes the images’ correlation structure again and can improve this AUC. Stationary textures sampled from zero mean and unequal covariance Gaussian distributions allow closed-form analytic expressions for the IO and optimal linear compression. Linear compression is a mitigation technique for high-dimension low sample size (HDLSS) applications. By definition, compression strictly decreases or maintains IO detection performance. For small quantities of training data, linear image compression prior to the sCNN architecture can increase AUC from 0.56 to 0.93. Results indicate an optimal compression ratio for CNN based on task difficulty, compression method, and number of training images.

A Survey of Rotation Invariant Texture Classification Methods

2002 ◽  
Vol 48 (3-4) ◽  
pp. 189-198
Author(s):  
Ramchandra Manthalkar ◽  
P K Biswas
Keyword(s):  
Texture Classification ◽  
Classification Methods ◽  
Rotation Invariant

Rotation invariant features for color texture classification and retrieval under varying illumination

2011 ◽  
Vol 16 (1) ◽  
pp. 69-81
Author(s):  
B. Sathyabama ◽  
M. Anitha ◽  
S. Raju ◽  
V. Abhaikumar
Keyword(s):  
Texture Classification ◽  
Rotation Invariant ◽  
Invariant Features ◽  
Color Texture

Efficient and Adaptive Rotation Invariant Wavelet Transform

2003 ◽  
Vol 01 (03) ◽  
pp. 353-372
Author(s):  
Chi-Man Pun
Keyword(s):  
Wavelet Transform ◽  
Texture Classification ◽  
Wavelet Packet ◽  
Sampling Rate ◽  
Discrete Wavelet ◽  
Wavelet Coefficients ◽  
Information Cost ◽  
Image Size ◽  
Rotation Invariant ◽  
Adaptive Wavelet

It is well known that the sensitivity to translations and orientations is a major drawback in 2D discrete wavelet transform (DWT). In this paper, we have proposed an effective scheme for rotation invariant adaptive wavelet packet transform. During decomposition, the wavelet coefficients are obtained by applying a polar transform (PT) followed by a row-shift invariant wavelet packet decomposition (RSIWPD). In the first stage, the polar transform generates a row-shifted image and is adaptive to the image size to achieve complete and minimum sampling rate. In the second stage, the RSIWPD is applied to the row-shifted image to generate rotation invariant but over completed subbands of wavelet coefficients. In order to reduce the redundancy and computational complexity, we adaptively select some subbands to decompose and form a best basis representation with minimal information cost with respect to an appropriate information cost function. With this best basis representation, the original image can be reconstructed easily by applying a row-shift invariant wavelet packet reconstruction (RSIWPR) followed by an inverse polar transform (IPT). In the experiments, we study the application of this representation for texture classification and achieve 96.5% classification accuracy.

Sign in / Sign up

Export Citation Format

Share Document