Editor, Based Image Display Adjustment With Scale Changes

Image editing is one of the new arising application fields in image processing. With the increase of the application area of image editing, it spans many subjects including computer vision, pattern recognition, artificial intelligence and image processing etc. Each year, there are a lot of publications appearing in the premier conferences and journals. Using artificial auxiliary function and automatic identification to edit or adjust the image, the aim of image editing is to meet all the needs for application requirements. Image editing is a subject which we need exploiting our great imagination to make innovations and practice. For the image editing covers a wild field, we just focus our research on the representative and hot areas of appearance editing and image resizing.Most of existing algorithms have many restrictions such as needing users to do interaction many times and processing time is too long etc. In this paper, after explored these main editing algorithms and its applications, we analyses the relation between the local and the whole data in the image, then present and implement a framework for accelerating the adjustment process of appearance editing and image retargeting. In particular, our contributions as follow:We propose a convenient user interface using strokes. In our framework, the user just indicates regions of interest by drawing some rough strokes. Then a general formulation is used to the initialize for different edit application. Our tools are also produce an adjustment method base on the color harmonization automatically, its maybe the prime exploration on automatic editing in future.We derive a hierarchical clustering method to group pixels into clusters using k-d tree in the adjustment of image appearance. Our edits propagating base on the cluster processing. In each cluster, we use the amount of user-edited pixels to measure its weight. The propagating process exploring a hierarchical interpolation method between clusters, which is guided by the general principle that similar edits should be applied to regions of similar appearance. For all the clusters are independent, the propagating process can be parallel, so it meets the requirement of real-time processing and high-resolution.Recent algorithms for image retargeting allow images to be resized to a new aspect ratio. Unfortunately, these approaches are extremely slow. In contract, we present an energy function for each patch, then using the function to construct our hshing algorithm-HashRetar for patch matching between images. The algorithm provides a hash-table for searching the matched patch in images. We demonstrate that HashRetar works on several appearance quantities and these tools is fast enough that the user quickly sees intermediate results.In conclusion, we present a novel method for adjusting image appearance and resizing the ratio of images. From the performance and quality of processed result, our system meets user requirements and improves user experience. Furthermore, we believe that our technique can be wildly used in art design and visual effects etc.