Search

Anamorphosis is related to the art that gives illusion (distortion) over the image or object when the viewer looks the original image or from the random viewpoint. This illusion effect is nullified when the image or object is viewed from the standard viewpoint. In digital imaging domain, anamorphic images (distortion) are generated by changing the pixel coordinates of the given image into new pixel coordinates, and this process typically increases the size of the original image matrix which results in gaps between the successive pixels. In anamorphic transform, the gaps between the successive pixel positions increases as we travel from viewing end to the tail end, and these gaps are filled by interpolating the successive pixels. The viewing angle, distance, height between the viewer and original image are responsible for the amount of distortion introduced in anamorphic images. The proposed work reveals that the rotation of the original image also plays an important role in creating distortion in anamorphic images. The proposed work uses the Kullback-Leibler distance (KLD) to measure the amount of distortion between the original and anamorphized images. A larger KLD implies that the original and anamorphic images are well separated i.e., a good anamorphism has achieved. It is found that the smaller viewing angle combined with the viewing distance and height achieves the larger KLD. The optimum anamorphic image generation is done at two levels: (i) Choosing an appropriate viewing angle, distance, and height, and (ii) Choosing the appropriate rotation of the image to be anamorphized. The former is optimized through KLD, and for the latter Shannon's entropy of the tail portion of the original image under different rotations is used. The proposed technique is tested on various images with different resolutions and is found to be working properly. [ABSTRACT FROM AUTHOR]

An image can be represented by different types of feature descriptors. Each descriptor holds specific details about the image. The image classification and retrieval applications use these extracted features to classify the query image and retrieve similar kinds of images presented in huge databases for the given query image. The discriminant feature in the form of a binary pattern available around each pixel of the image is extracted by taking the local pixel difference between each pixel and its neighboring pixels (i.e., sampling points) present in different radii. The discrete wavelet representation (i.e., multi-resolution) of the image has more information about the image compared to the image present in the spatial domain since each level of the multi-level decomposition discloses significant details about the image in separate channels. However, binary pattern extraction around each coefficient of the different levels of decomposed images lacks in producing the discriminant texture feature representation since it extracts a feature from each decomposition level at a time and concatenates them. Thus, the proposed work extracts the binary pattern around each coefficient of different levels of decomposed image from different radii. Then, the obtained binary patterns available in different decomposition levels are encoded to represent the discriminant texture feature around each pixel location. Consequently, the proposed work selects the multi-level decomposition based on the stationary wavelet transform, since it has the same resolution in each level of decomposition. The performance of the proposed feature descriptor is assessed over seven different sets of databases using image classification and image retrieval applications. Finally, the performance of the proposed work is compared with the state-of-the-art techniques involved in extracting spatial arrangement details of pixels available in the image. [ABSTRACT FROM AUTHOR]

The content-based image retrieval (CBIR) system accepts the input in the form of images and retrieves the relevant images from the database. The CBIR system automatically extracts the prominent key information from the image involved in the retrieval task. The color is one of the key information of the image and it is represented by dominant color descriptors (DCD). Here, similar colors get clustered and the mean value of each cluster represents the dominant color. The random number of unstable cluster formation in DCD alleviates the CBIR system performance. The proposed work has minimized the drawback of DCD by introducing seed points selection based on the mean, maximum and minimum value of the color pixels present in the image. Moreover, this work suggests the optimal cluster number by validating the different combinations of the proposed stable dominant color clusters. The retrieval precision of the proposed CBIR has improved since this work gives equal weight for both the dominant color and its occurrence probability in distance metric calculation. Finally, four standard datasets namely Wang's, Corel-10k, OT-scene, and Oxford flower are considered for evaluation, and it gives more number of relevant images compared to the state-of-the-art dominant color feature extraction techniques used on these datasets. [ABSTRACT FROM AUTHOR]