A Proposition of a Robust System for Historical Document Images Indexation

Abstract

Characterizing noisy or ancient documents is a challenging problem up to now. Many techniques have been done in order to effectuate feature extraction and image indexation for such documents. Global approaches are in general less robust and exact than local approaches. That's why, we propose in this paper, a hybrid system based on global approach (fractal dimension), and a local one, based on SIFT descriptor. The Scale Invariant Feature Transform seems to do well with our application since it is rotation invariant and relatively robust to changing illumination. In the first step the calculation of fractal dimension is applied to images, in order to eliminate images which have distant features than image request characteristics. Next, the SIFT is applied to show which images match well the request. However, the average matching time using the hybrid approach is better than "fractal dimension" and "SIFT descriptor" techniques, if they are used alone.

Deep detector classifier (DeepDC) for moving objects segmentation and classification in video surveillance

“…Zaghden et al . [39] proposed a method based on a global approach (fractal dimension), and a local one, based on the scale‐invariant feature transform (SIFT) descriptor for feature extraction to categorise ancient documents. In [40], the authors focused mainly on the characterisation of classes in images of ancient documents.…”

Characterization of ancient document images composed by Arabic and Latin scripts

“…The method proposed in [1] is based only on global approaches by extracting fractal dimensions of each connected components. In [2], the authors present a method based on global and local approaches for the indexation of historical document images problem. The comparison of pages of historical documents is studied here where in the first step, Zaghden proposes the reject of distant images classes by the comparison of their fractal dimensions.…”

“…The comparison of pages of historical documents is studied here where in the first step, Zaghden proposes the reject of distant images classes by the comparison of their fractal dimensions. In the second step the application of local features based on SIFT descriptors [2] was done, in order to retrieve the best matching images for the image request.…”

Automatic image annotation system using deep learning method to analyse ambiguous images

“…AIA models concepts with easily accessible pre-annotated photo sources. The trained model will then be used to complete incompletely labelled photographs or to classify unidentified photos [5]. Aerial photography is one of the subfields of remote sensing.…”

“…(4) Three-dimensional perspective Measurements in both the horizontal and vertical planes are possible because to the stereoscopic portrayal of the Earth's surface in aerial photographs. (5) Resolution: Images can be affected by electromagnetic waves with wavelengths which are beyond the range of the human eye's sensitivity. (6) Availability: Images may be captured at user-specified intervals and in a variety of sizes, providing a permanent record of any location.…”