In recent years, the resolution of display devices has been extremely increased. The resolution of video camera
(except very expensive one), however, is quite lower than that of display since it is difficult to achieve high spatial
resolution with specific frame rate (e.g. 30 frames per second) due to the limited bandwidth. The resolution
of image can be increased by interpolation, such as bi-cubic interpolation, but in this method it is known that
the edges of image are blurred. To create plausible high-frequency details in the blurred image, super-resolution
technique has been studied for a long time.
In this paper, we proose a new algorithm for video super-resolution by considering multi-sensor camera
system. The multi-sensor camera can capture two types video sequence as follow; (a) high-resolution with low
frame rate luminance sequence, (b) low-resolution with high frame rate color sequences. The training pairs for
super-resolution are obtained from these two sequences. The relationships between the high- and low-resolution
frames are trained using pixel-based feature named "texton" and stored in the database with their spatial
distribution. The low-resolution sequences are then represented with texton and each texton is substituted by
searching the trained database to create high-resolution features in output sequences.
The experimental results showed that the proposed method can well reproduce both the detail regions and
sharp edges of the scene. It was also shown that the PSNR of the image obtained by proposed method is improved
compared to the image by bi-cubic interpolation method.
It is important to estimate the noise of digital image quantitatively and efficiently for many applications such as noise
removal, compression, feature extraction, pattern recognition, and also image quality assessment. For these applications,
it is necessary to estimate the noise accurately from a single image. Ce et al proposed a method to use a Bayesian MAP
for the estimation of noise. In this method, the noise level function (NLF) which is standard deviation of intensity of
image was estimated from the input image itself. Many NLFs were generated by using computer simulation to construct
a priori information for Bayesian MAP. This a priori information was effective for the accurate noise estimation but not
enough for practical applications since the a priori information didn't reflect the variable characteristics of the individual
camera depending on the exposure and shutter speed.
In this paper, therefore, we propose a new method to construct a priori information for specific camera in order to
improve accuracy of noise estimation. To construct a priori information of noise, the NLFs were measured and
calculated from the images captured under various conditions. We compared the accuracy of noise estimation between
proposed method and Ce's model. The results showed that our model improved the accuracy of noise estimation.
We estimated and analyzed the reflectance spectra of human skin at each pixel taken by a multi-band imaging system, and compared it with the RGB three bands imaging system. The multi-band camera can capture more than three bands, and estimate the spectral reflectance using more than three eigenvectors of principal component. In this paper, we implemented a six bands camera system, in which color of the light source is modulated for rapid measurement. The system consists of the RGB three bands digital camera and two different light sources which are strobe light and strobe light with optical filter. We captured the skin twice using each light source respectively. As a result, we could obtain six color bands images. We estimated the spectral reflectance image based on Wiener estimation method, by using the captured multi-band image and the RGB image respectively. Then, we analyzed and evaluated two spectral reflectance images by using principal component analysis. The result shows that that the reflectance spectra estimated from three-band image do not have enough accuracy to analyze the skin color quantitatively.