Search

Distortion representation is the key to the rectification of distorted document images. The text-lines are considered to be one of the most significant features of the images, which are extensively used by a majority of rectification algorithms. However, it is quite a challenge to accurately extract the text-lines of document images with distortions and other disruptive factors, such as non-textural objects. In this approach, we present a general document rectification method based on local distortion representation that is depicted by text-features instead of the text-lines. Specially, firstly, according to the similarity of local distortion, we divide the document image into local blocks. Secondly, a text-feature is exploited to depict the warping distortion of each block by considering the skew angle. Then, the rectification problem is formulated utilizing a reverse strategy according to the text-features. Finally, a perspective distortion is restored by making use of random sample consensus. The proposed method is appropriate for document images of multi-column layouts, multi-type fonts and non-textural objects. Various experiments have demonstrated the flexibility and high performance of the approach.

Geometry images are important representations of 3D geometry models. Smooth geometry images contributes to low approximation error and high image compressibility. We present an adapted parameterization method to generate a smooth geometry image. It is quite challenging to directly modify the parameter domain to smooth geometry images. Our novel idea is that we use an indirect way to construct a resulting parameter domain according to the desired geometry image. We first move image pixels according to the current parameter domain to decrease the local linear error. Then we formulate a relationship between the moved image pixels and the current parameter domain. Finally, we use the relationship to update the parameter domain by restituting the image pixels to their original positions. The process will continue until the local linear error is less than a given threshold or the number of iterations is larger than a given threshold. Experimental results illustrate that geometry images generated by our method have low linear errors and low approximation errors under different sampling resolutions.