Image enhancement by nonlinear extrapolation in frequency space

Abstract

A technique for enhancing the perceptual sharpness of an image is described. The enhancement algorithm augments the frequency content of the image using shape-invariant properties of edges across scale by using a nonlinearity that generates phase coherent higher harmonics. The procedure utilizes the Laplacian transform and the Laplacian pyramid image representation. Results are presented depicting the power-spectra augmentation and the visual enhancement of several images. Simplicity of computations and ease of implementation allow for real-time applications such as high-definition television (HDTV).

IR Image Segmentation by Combining Genetic Algorithm and Multi-scale Edge Detection

“…Compared with the former segmentation algorithms this detection algorithm could not only extract edges of a target clearly but also the edges succession behaves better because of the "non-maximum restraint" for pixels gradient in course of optimization. From Fourier transformation frequency spectra Fig.3(a), Fig.3(b) and Fig.3(c) relative to Fig.2(b), Fig.2(c) and Fig.2(d) the effect of the Laplace-Gauss edge detection [7] could be evaluated. Their energy is basically along the main axis and lower inside quadrants with a great loss of details information.…”

Arbitrary Scale Image Enlargement with the Prediction of High Frequency Components

“…1 we can find that the LP stages of a step signal have the similar structure, and the difference is only that the slope is different in the neighborhoods of the zero-crossing point. Therefore, we can predict the high frequency stage Ä ½ by changing the slope of the stage Ä ¼ near the zerocrossing point [8]- [11]. However, [8]- [11] are limited to the representation of the integer pyramid stages and only have the ability of expanding an image up by a factor of two in size.…”

“…Therefore, we can predict the high frequency stage Ä ½ by changing the slope of the stage Ä ¼ near the zerocrossing point [8]- [11]. However, [8]- [11] are limited to the representation of the integer pyramid stages and only have the ability of expanding an image up by a factor of two in size. In many applications we also need arbitrary enlargement scales, such as 1.41 times and so on.…”

“…[6] uses the DCT iteration method for RE enlargement, and [7] uses the orthogonal wavelet transform method for RE enlargement. OIRE methods [8]- [11] are based on the Laplacian pyramid (LP) representation. They are called LP methods.…”

“…Utilizing the characteristic that the LP stages at different resolutions have the similar structure in the neighborhoods of the zerocrossing point, the LP methods can predict the unknown high frequency components to enhance the resolution of lowresolution images. Unfortunately, the LP methods of [8]- [11] are limited to the integer LP stages calculation, and thus they are only capable of expanding an image up by a factor of two in size (so called "zoom in"). In fact, we also need the enlargement scales as 1.41 times, 2.15 times and so on for many enlargement applications.…”