In this paper we present an algorithm for edge and corner detection in the greyscale and colour images using Fourier Amplitude Measure. We use the magnitude of the first Fourier basis to detect changes in the image intensity profile. We show the mathematical equivalence of the first Fourier co-efficient to that of the first order partial derivatives. Thus we interpret these coefficients as an estimate of the derivative measure and call them as Fourier Derivative Estimate. We then modify the first order derivative based method used for edge and corner detection using the Fourier Derivative Estimates (FDE) to produce robust edges and corner points. We compare the repeatability rate between FDE method and first order derivative methods in grayscale images. We then show examples where in, the generalization of this method to color images is found to be stable under illumination, blur, color contrast and affine variations, suggesting the utility of this method for image registration and mosaicing purposes.
In this paper we have proposed a method, called the Fourier Feature Extractor (FFE) that relies on orthogonal sinusoidal bases for feature extraction. The response generated on projecting image intensity windows onto these sinusoidal bases is used for extract features. We use energy of Fourier basis to capture the intensity change around pixel points. By interpreting the amplitude values of Fourier basis we distinguish different interest points like, edge-point, corner-point, T-junctions. The method offers the flexibility to choose different kinds of interest points depending on the choice of basis function and energy values. The feature points detected are found to be geometrically stable under different transformations. The Fourier measure assigned during extraction can be used for matching of feature points between images and this make registration efficient.
The mosaicing system proposed in this paper consists of a temporal segmentation module and a motion estimation module carried out in the MPEG (1&2) domain. The input MPEG video is temporally segmented into shots by the segmentation module and each shot is mosaiced separately. The output of the system is a set of mosaics, each of which captures the panorama of each shot uniquely. The camera motion parameters are computed for each shot using the MPEG motion vectors; frames from a shot are then aligned and integrated into a static mosaic for each shot. The proposed system is 200-300% faster in executing the registration step for our input video sequences.
A novel approach for image mosaicing in the JPEG compressed domain is presented in this paper. This technique employs the Hausdorff Distance Metric (HDM) to compute the regions of overlap between two JPEG images. The DCT blocks of the two overlapping images having significant activity are identified using a variance measure and the HDM metric is employed directly between these DCT blocks to estimate the translation parameter. The results obtained demonstrate reduction in the time taken for the registration step by a factor of 10-30, as compared to a traditional feature-based approach in the uncompressed domain.
Video Mosaicing is the process of obtaining single unified picture from multiple frames of a video sequence. It involves finding the geometric transformation parameters between frames and aligning them to form a single image representing content of full video. The proposed mosaicing method takes advantage of the information encoded by MPEG in the form of motion vectors, DCT blocks and the error information for generating the mosaics. Only the motion vectors corresponding to blocks of high activity region are considered for finding affine parameters. The feasibility of the method has been tested on panorama views from various MPEG sequences. The performance of proposed method was evaluated against the existing uncompressed domain methods.
Traditional feature extraction techniques like the KLT, Harris and Wavelet work only in the uncompressed domain. Hence an additional step of decompression is required before any of them could be applied. We propose a two-level technique for extracting high-level feature points directly from JPEG compressed images. At the first level, the Discrete Cosine Transform (DCT) blocks having high activity content are filtered using a variance measure. At the next level, a DCT block centered at every pixel present in the filtered block is constructed from the neighboring DCT blocks. Feature points are then selected by analyzing the AC coefficients of the DCT block centered about it. The proposed method is simple and efficient. The extracted feature points were found to be rich in information content, which could be used for image registration. The results of this technique showed almost the same amount of repeatability between two images with 60% to 70% overlap, when compared with techniques available in the uncompressed domain. The features thus extracted can directly be used to calculate the motion parameters between two images in the compressed domain.