This action might not be possible to undo. Are you sure you want to continue?

# Abstract

Usually there are differences between a recorded color image and the direct observation of the same scene. One of the big differences is the fact that the human visual system is able to distinguish details and vivid colors in shadows and in scenes that contain illuminant shifts. In this project we examine the performance of an algorithm called A Multiscale Retinex with Color Restoration (MSRCR), presented in [1]. This algorithm tries to imitate human visual “computation” while observing scenes that contains lighting variations. MSRCR is an extension of a former algorithm called Single Scale center/surround Retinex (SSR) [2]. MSRCR achieves simultaneous dynamic range compression, color consistency and lightness rendition. In this project we look for the optimal value of each parameter of the SSR and the MSRCR. The MSRCR algorithm was compared to other two, well known, image enhancement methods – gamma correction and histogram equalization. Results indicated that the performance of the Multiscale Retinex is consistently good, while the performance of the others is quite variable.

Basic Approach

The general mathematical formulation of the center/surround Retinex is

ˆ R i ( x , y) = log I i ( x , y) − log[ F( x , y) ∗ I i ( x , y)] ˆ where R i (x, y) denotes the Retinex output, Ii(x,y) the image distribution in the i'th color spectral band, “*” the convolution operation, and F(x,y) the surround function

F( x , y) = Ke

−r 2 / c 2

where c is the Gaussian surround space constant and K is selected such that The MSR output is then simply a weighted sum of the outputs of several different SSR output.

ˆ ˆ R MSR i = ∑ w n R n ,i

n =1 N

∫∫F( x, y)dxdy

=1

ˆ where N is the number of scales, the i'th component of the n’th scale, R MSR the i'th spectral component of the MSR output, and wn the weight associated with the ˆ ˆ n’th scale. The only difference between R(x, y) and R n (x, y) is that the surround function is now given by

ˆ R n ,i

i

Fn ( x , y) = K n e − r

2

/ c2 n

The color restoration method for the MSR is given by S C i ( x , y) = log[ αI i ( x , y)] − log ∑ I i ( x , y) i =1 where α is a constant parameter of the color restoration function. The MSRCR is given by

ˆ ˆ R MSRCRi ( x , y) = C i ( x , y) R MSRi ( x , y)

**The final version of MSRCR can be written as ˆ R MSRCRi ( x , y) = G[C i ( x , y)∑{log I i ( x , y) − log[ I i ( x , y) ∗ Fn ( x , y)]} + b]
**

n =1 N

**where G and b are the final gain and offset values respectively.
**

1. Daniel J. Jobson, Zia-ur Rahman, Glenn A. Woodell; “A Multiscale Retinex for Bridging the Gap Between Color Images and the Human Observation of Scenes”, IEEE Transaction on Image Processing, vol. 6, no. 7, July 1997.

2. Daniel J. Jobson, Zia-ur Rahman, Glenn A. Woodell; “Properties and Performance of a Center/ Surround Retinex”, IEEE Transaction on Image Processing, vol. 6, no. 3, March 1997.

. In order to ease the research and to make it more convenient to get the visual results. we also examine the following parameters of the MSRCR algorithm: • the combination three scales (small. intermediate and large) of the SSR functions • the weights associated with the scales • the color restoration constant. Results Figure 2 illustrates the MSRCR performance. The MSRCR algorithm was also tested on some images with familiar colors and no strong lighting defects and the results showed that the MSRCR does not introduce significant distortions into images without lighting variations. we developed a graphic user interface (GUI). Tools The algorithm was realized as a MATLAB program. As seen the MSRCR algorithm achieves graceful dynamic range compression together with tonal and color rendition. Figure 1: Graphic User Interface for the MSRCR algorithm. The GUI allows the user to change the MSRCR constants before running the algorithm and to compare the resulting performance to the performance of a standard brightness correction (gamma correction). see Figure 1.Project Goal • • • • In this project we examine several aspects concerning the SSR: the form of the surround function the placement of the log function the scale constant for the surround function the treatment of the Retinex result before display In addition.

Tests comparing MSRCR to other image enhancement methods: gamma correction and histogram equalization indicate that the performance of the Multiscale Retinex is consistently good. Histogram equalization may produces artifacts.original images. This includes clipping some of the lowest values in order to get better color rendition. c=80. exponential and Gaussian are quite similar. This range of C3 is quite close to the range presented in [2] ( 50 ≤ C 3 ≤ 100 ). • The Retinex output has to be processed before displaying. • The MSR is composed of three SSR functions with small(c=80). though sometimes its dynamic range compression is better than the MSRCR’s. in which the Gaussian’s performance was significantly better. Gamma correction produces unsharp masking that damages the color rendition and blurs the details. Also we would like to thank Johanan Erez for his attention and help during the whole project. Acknowledgments We would like to thank our supervisor Dr. color consistency and tonal rendition. The gain and offset constants do not vary either from image to image or between color bands. This result stands in contrast to the results presented in [2]. while the performance of the others is quite variable. c=250. • The placement of the log function should be after surround formation.Figure 2: The performance of the MSRCR: top row . . We chose larger constant values because small values produce desaturation of colors in the final image. except for scenes that contain violation of the gray-world assumption. The Gaussian function was chosen as the form of the surround function because of compliance considerations and due to its small width in comparison to the other two. The color restoration function produces good color rendition even when a gray-world violation occurs. intermediate (c=120) and large (c=250) scale-constants. The scale constants that were chosen in [1] are: c=15. Doron Shaked for his support and guidance throughout this project. bottom raw – MSRCR Conclusions • The performance of the three surround functions: inverse square. This project was supported by the Ollendorff Research Center Fund. • The scale constant of the Gaussian function should be 80 ≤ C 3 ≤ 120 in order to get a satisfying image quality. This combination allows the MSR to synthesize dynamic range compression. but occasionally it produces better color rendition than the MSRCR.