In converting line image data of a minute shape portion into polygonal line data, a pattern recognition apparatus is designed so as to determine polygonal line data from line image data by conventional polygonal line conversion processing technique, to presume an optimum shaping for for a shape formed by a group of the polygonal line data from the relationship with the drawing therearound, and to judge whether a path of the shaping form passes on the line image at a given constant percentage or larger. This permits an approximation of a minute shape portion likely to undergo an influence from digital noise to polygonal line with a high reliability.

A block copy image is divided into several regions R1 and R6, to which different system colors are allocated. A subject region PR including a line break CP1 is specified by an operator. Run-length data of a run-length less than or equal to a predetermined threshold value is automatically detected out of various run-length data representing the subject region PR. White/black designation data defining the color of an image in the run-length data is changed into the color of a linework. The break CP1 in the linework is thus reconnected without operator's instruction. Alternatively, the subject region PR is divided into two portions. The two portions are further divided into separate areas to which different system colors are allocated. This enables an operator to easily detect the position of a line break because the linework elements separated by a break have different colors. The break is then reconnected according to operator's instruction.

The labelling circuit of this invention comprises first and second table memories, which designate when pixels assigned with different label values are of the same linked components and whose contents are updated based on linked information produced by a label extracting circuit. The labelling circuit further comprises first through third selectors. The first and second selectors select the label value of an intermediate processed image previously produced by the label extracting circuit, or one of label values of new linked information produced by the label extracting circuit, as an address for the first and second table memories. The third selector selects one of output data of the first and second table memories as output image data.

A boundary line detection apparatus and method and an image processing apparatus and method as well as a recording medium by which the jaggy in the proximity of a boundary line is removed without increasing the sampling rate to remove are disclosed. The differences between pixel values between a noticed pixel and pixels in the proximity of the noticed pixel are compared with each other to detect whether or not a boundary is present in the proximity of the noticed pixel. Then, a length, a direction, and a start point and an end portion of the boundary line are determined. Then, the distance between the noticed pixel and the boundary line is calculated, and the pixel value of the noticed pixel is weighted using the distance. Then, the pixel obtained by the weighting is outputted.

A method for finding a predefined object in an image includes selecting a line belonging to the object, the line having a known stroke width and defining a maximum width and a minimum width that together define a range of widths therebetween that contains the stroke width. A feature in the image is found having a feature width within the range, and the feature is processed to determine whether it is a part of the object.