Abstract: When a scan line is interpolated, pixels on scan lines above and below the scan line which are referenced is interpolated in the horizontal direction. These pixels on the upper and lower scan lines are divided into blocks. Correlations between the upper and lower blocks with respect to a target pixel to be interpolated are calculated. Based on an evaluating function which is formed based on the correlations but is corrected with reference to the distance (or the direction) from the target pixel to each of the blocks, an optimal interpolating direction is selected. In addition, the difference in the pixel data between the interpolated pixel and each of the pixels around the interpolated pixel is checked in order to find interpolation errors. When an interpolating pixel data is determined to be erroneous, the pixel is interpolated using the pixels of data above and below the pixel.

Abstract: An image encoder for encoding the image data so as to make the image quality of a selected area better than that of the other areas without increasing the amount of data is disclosed.

Abstract: The prediction error value calculated by a motion estimating portion is used for selecting the frame or field orthogonal transform coding mode. The frame prediction error value EFr and the field prediction error value EFi are calculated by the motion estimating portion and entered into a comparator of an orthogonal transform mode selecting portion which in turn compares them with each other and generates an orthogonal transform mode selecting signal TC. The use of the prediction error values for selecting the orthogonal transform mode makes it possible to reduce the size of the necessary hardware and saves much time of data processing necessary for judgment of the orthogonal transform mode selection.

Abstract: The prediction error value calculated by a motion estimating portion is used for selecting the frame or field orthogonal transform coding mode. The frame prediction error value EFr and the field prediction error value EFi are calculated by the motion estimating portion and entered into a comparator of an orthogonal transform mode selecting portion which in turn compares them with each other and generates an orthogonal transform mode selecting signal TC. The use of the prediction error values for selecting the orthogonal transform mode makes it possible to reduce the size of the necessary hardware and saves much time of data processing necessary for judgment of the orthogonal transform mode selection.

Abstract: In an image encoding apparatus, image information is quantized and variable-length encoding is performed thereon so that the data are stored in a buffer. Thereafter, the data in the buffer are transmitted to the outside at a fixed data rate. At this time, a controller monitors a number of occupied bits in the buffer, and when the number of occupied bits is not more than predetermined thresholds, the controller switches the output of the apparatus to a dummy data generator. At this time, a parameter which is used for rate control is initialized, and dummy data which can be eliminated at a decoding circuit are outputted from the dummy data generator. Furthermore, bit allocation is changed so that a sum of the number of occupied bits and the bit allocation to a frame, which is determined before encoding, does not exceed a fixed amount. Moreover, when the number of occupied bits exceeds the fixed value, a quantization step size is changed to a greater value.

Abstract: A motion vector detecting circuit is for detecting a motion vector in moving picture coding using motion compensation prediction in the unit of block composed of L pixels by block matching between a block in a prediction target image and a candidate block in a reference image. The motion vector detecting circuit comprises a calculating section for calculating a first prediction error between a block in a prediction target image and a candidate block in a reference image using L1 (<L) pixels in block, comparing the first prediction error thus calculated with a threshold, and calculating a second prediction error between a block in the prediction target image and a candidate block in the reference image using L2 (L1<L2.ltoreq.