Teena Sharma's research while affiliated with Indian Institute of Technology Guwahati and other places

This chapter presents the introduction of an Adaptive Interval Type-2 Fuzzy Filter. This filter is presented as one of the artificial intelligence agents for the preservation of naturalness in non-uniformly illuminated images. The filter estimates the coarse illumination within the input image, and subsequently, undergoes further processing to obtain the reflectance and refined coarse illumination. The estimated coarse illumination plays a pivotal role in preserving naturalness by addressing uncertainties arising from grayness ambiguities in homogeneous regions and spatial ambiguities at edges. This chapter also delves into various applications to showcase the efficacy of the filter.

This chapter presents a single image dehazing method employing a Type-2 membership function based similarity function matrix. The approach involves estimating the depth map and global atmospheric light from the observed hazy image. Subsequently, the estimated depth map is used to generate the true scene transmission. Finally, the observed hazy image undergoes dehazing through the atmospheric scattering model, utilizing the scene transmission and global atmospheric light. The chapter showcases an extensive set of experimental results, demonstrating the superior performance of image dehazing using the Type-2 fuzzy approach.

In addressing hazy conditions, techniques for enhancing image quality predominantly center on estimating the transmission map to restore a haze-free image. However, the conventional transmission map estimation function, based on the atmospheric scattering model, tends to suffer from issues such as over-saturation and over-dehazing. This is primarily attributed to its limited variability with respect to the scene depth. This chapter introduces an approach by presenting a modified transmission map estimation function. This function incorporates a saturation component as an additional haze-relevant feature. The inclusion of this saturation component aims to mitigate problems associated with over-saturation and over-dehazing, thereby enhancing the effectiveness of image quality improvement techniques in hazy conditions.

In the previous chapter, the unsupervised image dehazing method was explained, which deals with over-saturation challenges. This chapter delves into a detailed exploration of the Compact Single Image Dehazing Network (CSIDNet) as an effective solution for enhancing outdoor scenes through deep learning-based single image dehazing. CSIDNet outperforms various state-of-the-art dehazing models and also establishes itself as a compact model with minimal resource requirements. The enhanced scene images produced by CSIDNet successfully strike a balance between speed and accuracy. The chapter presents the results of CSIDNet highlighting the potential application of CSIDNet in critical and real-time scenarios.

This chapter introduces an image enhancement approach utilizing the Z-score for images captured under hazy and non-uniform illumination conditions, with a focus on applications in multi-media. The presented approach involves estimating scene transmission through a Z-score based weighting function and utilizing global atmospheric light for image dehazing. In contrast, the method equalizes the illumination channel by employing the Z-score weighting function for non-uniformly illuminated images. The chapter includes a comparative analysis to demonstrate the quantitative and visual effectiveness of the Z-score approach.

This paper proposes an algorithm to restore low-light images by enhancement of its inverted illumination map using dehazing. The proposed algorithm follows the Retinex theory and decomposes the low-light input image into reflectance and illumination components. The illumination map is then inverted and enhanced using dehazing-type algorithm and gamma correction. The enhanced map is again inverted back to obtain the enhanced illumination, and finally, it is combined with the original reflectance to output an enhanced image. The enhanced images obtained using the proposed algorithm have improved contrast and visual quality. The proposed method is analyzed qualitatively on various standard images used in literature and compared quantitatively with several benchmark techniques. It is found that the proposed algorithm is superior over various other benchmarks.KeywordsLow-light image restorationIllumination map enhancementRetinex theoryGamma correction

Image denoising is one of the preliminary steps in image processing methods in which the presence of noise can deteriorate image quality. This paper presents an improved two-stage fuzzy filter for filtering salt and pepper noise from the images to enhance the image quality. In the first stage, the pixels in the image are categorized as good or noisy based on adaptive thresholding using type-2 fuzzy logic with exclusively two different membership functions in the filter window. In the second stage, the noisy pixels are denoised using modified ordinary fuzzy logic in the respective filter window. The proposed filter is validated on standard images with various noise levels. The proposed filter removes the noise and preserves beneficial image characteristics, i.e., edges and corners at higher noise levels. The performance of the proposed filter is compared with the various state-of-the-art methods in terms of peak signal-to-noise ratio and computation time. The average PSNR values for the noise percentage 20%, 50% and 80% are the 37%, 31% and 27%. To show the effectiveness of statistical filter tests, i.e., the Friedman test and Bonferroni−Dunn (BD) test are also carried out, which ascertain that the proposed filter outperforms in comparison to various filtering approaches.