Image Enhancement Techniques - kovimesterr/SSIP2013 GitHub Wiki

One can find a vast range of enhancement techniques in the literature among which the most known techniques for contrast enhancement are listed below:

• Adaptive Median Filter
• Histogram Equalization
• Histogram Modified Local contrast Enhancement
• Contrast Limited Adaptive Histogram Equalization (CLAHE)
• Fuzzy contrast Enhancement

The above algorithms are particularly designed to enhance the contrast. As for the noise removal, since we do not have any knowledge of the type of the noise in the mammograms, we can not apply the classical noise removal approaches with known noise such as Gaussian. However, assuming that the degradation of mammograms can be represented by a convolution, we try Blind Deconvolution, as a tool for investigating whether we can find the particular kernel representing the possible PSF (point spread function) of this convolution. We will also discuss the result of our experiment at the end of this section.

Adaptive Median Filter

The main idea of the median filter [1] is to run through the image pixel by pixel, replacing each pixel with the median of neighboring pixels. The pattern of neighbors is called the "window", which slides, pixel by pixel, over the entire signal. The median filter is a nonlinear filter which , under certain conditions, can preserve edges , and it is often used to remove noise as a pre-processing step to improve the results of later processing (for our example, cancer segmentation).

Histogram Equalization

This technique [2] when applied on an image, adjust the contrast of the image using the image's histogram. The method usually increases the global contrast of images, especially when the usable data of the image is represented by close contrast values. This allows for areas of lower local contrast to gain a higher contrast. Histogram equalization accomplishes this by effectively spreading out the most frequent intensity values. The method is useful in images with backgrounds and foregrounds that are both bright or both dark and this is particularly true for the mammograms.

Histogram Modified Local contrast Enhancement

HE uniformly distributes the output histogram by using cumulated histogram as its mapping function. However it produces over enhancement in the output image which leads to loss of more local information in the original mammogram. One more problem with HE is its large backward difference values of mapping functions and the contrast enhancement potential should be enriched without loosing the fine details in the mammogram image. In order to lessen the level of enhancement that would be obtained by HE, the input histogram can be altered so that the modified histogram is closer to a uniformly distributed histogram. HM-LCE method incorporates a two stage processing both histogram modification and local contrast enhancement technique. The main objective of this method is to find a modified histogram that is closer to uniform histogram and to make the difference between modified and input histogram small, which in turn increases the potentiality of image contrast enhancement and resultant image would be the more relevant to the input image.

Although the global approach for image contrast enhancement is suitable for some cases, there are situations in which it is necessary to enhance local details in the mammogram image. The number of pixels in this area may have negligible influence on the computation of the global transformation. The solution is to device transformation function based on gray level distribution or other properties in the neighborhood of every pixel in the image. This method of approach is called local contrast enhancement.

We have already implemented this method, but results were not the same as in [3]. Namely, in the first step we could get back the same image as in [3], but after LCE the result has not really changed. The implementation of this function is available in the project directory.

Morphological Operators

Top-hat Morphological Processing

Top-hat morphological processing uses gray scale opening to extract regional maxima or objects which differ in brightness from the surrounding background in images with uneven background intensity [4] [5]. The high intensity regions, i.e., the features that cannot accommodate the SE are removed by performing a structural opening. The features removed by opening are emphasized by subtraction of opened image (OI) from the original image (I), which yields a top-hat transformed image.

Top-hat morphological processing returns an image of objects from the input image which are brighter than their surroundings and smaller then the SE.

H-dome Morphological Processing

H-dome morphological processing is yet another algorithm used to extract regional maxima based on grayscale reconstruction. Grayscale reconstruction [1] is a morphological operator based on two images instead of a single image and a SE. In grayscale reconstruction the original image is the mask image and marker image is created by subtracting a constant value h (here we used 60) from each pixel value in the mask image. The process of grayscale reconstruction can be viewed as repeated dilations of marker image under mask image:

where pI is the grayscale reconstructed and I is the original image. One can see an example on the next figure.

**Contrast Limited Adaptive Histogram Equalization (CLAHE) **

CLAHE [6] is a technique used particularly to improve contrast in images. It differs from ordinary histogram equalization in the respect that the adaptive method computes several histograms, each corresponding to a distinct section of the image, and uses them to redistribute the lightness values of the image. It is therefore suitable for improving the local contrast of mammogram images and bringing out more detail to the attention of doctors.

Fuzzy Contrast Enhancement

Fuzzy image enhancement is based on gray level mapping into a fuzzy plane, using a membership transformation function. The aim is to generate an image of higher contrast than the original image by giving a larger weight to the gray levels that are closer to the mean gray level of the image than to those that are farther from the mean. We tried to implement an interesting work which uses the same concept for enhancing the contrast of mammograms [7], but unfortunately due to limited time, we were not able to fully implement this algorithm and include its final result in this report. In fact the algorithm presented in the original work, to the best of our knowledge, is lacking some details and the enhanced image resulting from our implementation is clearly wrong for a set of pixels (as can be seen on one set of result in Fig )Therefore for the rest of this report, we skip further explanation of the algorithm and elaborate more on the blind deconvolution and the final experimental results.

**Blind Deconvolution **

Blind deconvolution [8] is a deconvolution technique that permits recovery of the target scene from a single or set of "blurred" images in the presence of a poorly determined or unknown point spread function (PSF). Regular linear and non-linear deconvolution techniques utilize a known PSF. For blind deconvolution, the PSF is estimated from the image or image set, allowing the deconvolution to be performed. Researchers have been studying blind deconvolution methods for several decades, and have approached the problem from different directions.

In our experiment we just test this method on synthetic images to examine its effects. During these tests we applied the transformation on images which was blurred with different size of kernel functions. Then we used blind deconvolution to check the processed image and the reconstructed PSF functions. One can see an example on the next figure.

As we can see that, the resulting images are degrading as we are increasing the bluring kernel. By this reason, this method generates a lot of artifact near the edges. Additionally, the details have bee lost. It is a bad effect of deconvolution process which is caused by the unknown kernel and its bad approximation. Further research is required on estimating the PSF from the mammogram images.

References

• [1] Hwang, H.; Haddad, R., "Adaptive median filters: new algorithms and results," Image Processing, IEEE Transactions on
• [2] Russ, The Image Processing Handbook: Fourth Edition, CRC 2002, vol.4, no.4, pp.499,502, Apr 1995
• [3] M. Sundaram, K. Ramar, N. Arumugam, G. Prabin, Histogram Modified Local Contrast Enhancement for Mammogram Images, Applied Soft Computing, vol. 11, 2011, pp. 5809-5816.
• [4] L. Vincent, Morphological Grayscale Reconstruction in Image Analysis: Applications and Efficient Algorithms, IEEE Transactions on Image Processing, Vol. 2, No. 2, pp. 176-201.
• [5] H. S. Jagannath, J. Virmani, V. Kumar, Morphological Enhancement of Microcalcifications in Digital Mammograms, Journal of the Institution of Engineers (India) Series B, Springer, 2012, pp. 163-172.
• [6] Pizer, S.M.; Johnston, R.E.; Ericksen, J.P.; Yankaskas, B.C.; Muller, K.E., "Contrast-limited adaptive histogram equalization: speed and effectiveness," Visualization in Biomedical Computing, 1990., Proceedings of the First Conference on , vol., no., pp.337,345, 22-25 May 1990
• [7] Hassanien, AboulElla. "Fuzzy rough sets hybrid scheme for breast cancer detection." Image and vision computing 25.2 (2007): 172-183.
• [8] E. Lam; J.W. Goodman (2000). "Iterative statistical approach to blind image deconvolution". Journal of the Optical Society of America A 17 (7): 1177–1184.