【 Weak current College 】 one splice elimination method

Image mosaic technology (mosai) is image fusion, generally refers to the same type image fusion. He took a number of sites have overlapping information part of the image link together, get a complete, larger image, and remove the redundant information. Image mosaic technology application is extensive. For example, virtual reality panorama display and remote sensing image processing and other fields, have a wide range of applications. Image mosaic of the evaluation criteria are the mosaic of images, not only has good visual effects, but also as much as possible to keep the image spectral characteristics. Popular to say, that is, the more mosaic image of the "seamless", the effect is better. Of course, here's the "seamless", is not an absolute sense, but the human eye resolution within the "seamless".

Normally, when an image mosaic, mosaic borders, inevitably generates mosaicking. This is because the two sites to mosaic image in grayscale nuances will cause obvious seam. But in the actual imaging process, such nuances is difficult to avoid. Therefore the difficulty of image mosaicking is accurate for the position between the image and the relationship between two or more images smoothly linking together, get a global image. The basic idea of this article is to break through in the past, looking for stitching line, as long as you find the best stitching points points do a straight line as a mosaic of irrationality, but a closed values, in close range of values for each connection point, bring these points into polylines as stitching wire for concatenation.

Mosaicking elimination method

Traditional stitching sewing eliminate many of the methods, which use more methods; median filter method, the method using Wavelet transform, the weighted average method, etc.

Median filtering method to eliminate mosaicking

Median filtering method is the area around the joints for median filtering. In and around gray value difference larger pixel and the surrounding pixel closest value, thereby eliminating the light intensity of discontinuity. Median filter treatment joints near the Strip. This method is fast, but the quality in General. Smooth results will make the resolution of the image, the image detail can't tell, resulting image blur.

Application of wavelet transform approach to eliminate mosaicking

Wavelet transform method is more commonly a way he makes full use of multi-resolution Wavelet transform features, very good solution to the problem of seam stitching images. The principle is: as the wavelet transform a bandpass filter properties, in the different scales of wavelet transformation component, actually occupy a certain amount of bandwidth, the larger the scale j in the case of the higher the frequency, so each Wavelet component has a bandwidth less, put to the concatenation of the two images by Wavelet Decomposition method will they break down into different frequency of Wavelet Decomposition, as long as fine enough, wavelet component of bandwidth would be small enough. And then at different scales, select different stitch width, the 2 images at different scales of wavelet component first mosaic down, and then use the restore program to restore the entire image. This image is a good balance between clarity and smoothness 2 requirements. However, there is a drawback to using Wavelet transform, as Wavelet transform algorithm is more complex, the need in the wavelet domain for stitching treatment within the first, in the calculation process involves a large number of floating point arithmetic and boundary handling issues, to the actual production of bulk image processing computer memory overhead is high, and the processing time is longer, the concatenation is slow.

Use weighted smoothing methods to eliminate mosaicking

In practice, using more method also has the area of overlap is weighted smoothing method. This way of thinking is: image overlap area pixel point gray value Pixel consists of two images in grayscale values corresponding to the point and the weighted average RPixel LPixel, namely: Pixel-kXLPixel + (l-k) XRPixel where: k is the gradient factor that satisfy the condition: o k < < 1, in the area of overlap, follow the image from left to right in the direction of the image, the gradient from 1 k to 0, which implements the overlap area is on the left overlapping area slowly transition to right overlapping area of smoothing the concatenation.

Looking for the best stitching lines, using a sliding window in the image overlay district-by-row select grayscale values difference the smallest pixel as the best stitching points. However, if you follow this stitching point selection method, a new problem, often occur downlink mosaic point location differs from the more distant phenomenon, mosaic and down after sometimes because of larger differences between grayscale and caused new seams. To avoid this kind of phenomenon, not only adjacent splicing point gray value differences, but also to consider the location of adjacent stitching points should not be too far away. This introduces an aperture value T, select the best stitching points limit the aperture value. In addition to the first row and press the gray value difference between the minimum principles, each row of stitching points from a selected area selection: that is, the preceding line selected stitch point with columns and to the point as the center around the width is in the area of T. In this area, select one of the best stitch points. Each row of stitching elected after connection to one splice line, it is conceivable that this mosaic cord may be a polyline. In this way, because the rows are selected provisions of the neighborhood in grayscale difference the smallest point as mosaic phenomenon, the seam will be a big difference. At the same time, the T value and cannot be selected obtained is too large, should choose between 1 a 5 is preferred. Find out the best stitching sewing, as previously described weighted smoothing on the area of overlap and then make the transition, the great image quality.