Fuzzy automatic contrast enhancement based on fuzzy C-means clustering in CIELAB color space

Some traditional methods for image contrast enhancement are based on histogram equalization, which however has the drawbacks of producing visual artifacts or excessive image strengthening due to improper settings of enhancement parameters or non-smooth color adjustments in different color spaces. A novel method of Fuzzy Automatic Contrast Enhancement (FACE) is presented in this paper. FACE first performs a fuzzy clustering method to segment an image while the pixels with similar colors in the CIELAB color space are classified into smaller image clusters with similar characteristics. The pixels in each group are then spread out away from the center of the belonging cluster in the RGB color space in order to enhance the image contrast but keeping the similarity of pixel colors in the same cluster. A universal contrast enhancement variable (UCEV) was defined and optimized to maximize the image randomness (i.e. entropy of the image) in order to automatically enhance the image contrast. A more uncongested distribution of the image pixels ensures a greater image contrast. The proposed entropy-maximization process is capable of improving the image quality without any human-defined control parameters. The fully automated image enhancement process intelligently clusters the pixels with similar color characteristics and is general for the contrast enhancement of images in various color distributions. Many images with different color distributions were tested and the results showed that FACE is capable of avoiding visual artifacts and excessive strengthening. Compared with the traditional histogram equalization method, the proposed method shows higher effectiveness in contrast enhancement and performs better in retaining the colors of the original images.