In this paper, we propose a method which can morph a high free virtual viewpoint image keeping the epipolar geometry relations based on the View Morphing technique. It works just by the images coming from two cameras with same parameter without 3D reconstruction or camera calibration. Experiments confirm that the morphed view based on the two parallel view points can produce any satisfied image at any high degree position in the space.

View full abstract

In this paper, we described the algorithms of a curve fitting by use of spline and discussed the accuracy of the fitting in order to express hand-written characters with less number of knots. For the criterion of fitting, we applied the residuals, by which we determined the knots adaptively and iteratively. A method was considered to reduce the number of knots, and the method to estimate the residuals is proposed. The results of the examinations indicate that the proposed algorithm is the effectual method in knot reduction of written characters. The features of this method have the good accuracy of the fitting and the capability of making the fitting value at the origin coincide with the actual one so that the property of the characters is satisfied. A few examples of the curve fitting are illustrated in the paper.

View full abstract

This paper reports abstracts of the two methods already proposed separately, the information embedding method into error-diffused images and the block coding method of error-diffused images, and reports the results of the evaluation tests which combined the two proposed methods. Using the Cyan component of the test image N1, the evaluation test showed that the value of 64 is best for the tolerance, which is the tolerance value of the threshold value for binarizing an image, from the view points of information embedding size, compression ratio, and image quality. Therefore, degradation of image quality can be kept small for the variable sizes of information embedded up to the maximum size, and the compression ratio by the Jarvice error-diffusion method can be improved. For example, when the Floyd error-diffusion method is used, the effective experimental results show that the compression ratio is 46.8% and the MAE is 4.6 without information embedded, but the compression ratio is improved to 44.1% and the MAE degrades 1.4 to 6.0 with information of the maximum size which corresponds to 58% of the total pixels embedded.

View full abstract

A simple method is presented for generating a hybrid image which is perceived as one image when seen up-close while another image is visible from far away. We generate a far-away image by decomposing one image into blocks and quantizing it with an error diffusion dither. An up-close image is produced from another image by subtracting its mosaicking image from it. A hybrid image is generated by addition of far-away image and up-close one. We experimentally verified that our method is simpler than the method of Oliva et al. and gives a hybrid image with higher separability between two images.

View full abstract