Content-Based Image Retrieval with Cellular Automata - Cellular Automata in Image Processing and Geometry - page 145

m − 1 n 1

j = 0 (

−

2

C i , j − R i , j )

∑

i = 0

MSE ( C , R )= .

mn

Let p represent the maximum possible pixel value (the peak signal) in an image
(for example, 255 in a black and white image). Then the PSNR is given by

p 2
MSE

) .
The higher the peak signal to noise ratio, the better the technique. An important
issue in reaching the final denoised image, is the number of evolutions that the CA
must execute. If too many evolutions are performed, the quality of the image can
degrade noticeably. For a general good quality image with only some noise, between
one and five evolutions are usually sufficient.
Various authors offer comparisons of the CA-based methods and traditional
methods [6, 14, 16, 27, 31] for noise reduction.
Given a denoised image, the next step in image retrieval is to find the feature
vector based on shape, colour and/or texture [10].

10 log 10

( C , R

Fig. 8.1 Images with noise before and after the majority update rule CA filter was applied