Seminal papers in Image Processing

From Wasteland

• K. Pearson. "Mathematical contributions to the theory of evolution - Regression, heredity, and panmixia", Philosophical Transactions of the Royal Society of London, Series A, Containing papers of a mathematical or physical character, vol. 187, pp. 253--318, 1896.

First mathematically rigorous formulation of the concept of correlation between two variables. This is where Pearson's linear coefficient was defined (even though he called it the Galton function or coefficient of correlation).

• J. Cohen, "A coefficient of agreement for nominal scales", Educational and psychological measurement, 20(1):37-46, 1960.

Definition of Cohen's kappa coefficient, the first measure of interobserver agreement.

• H. Akaike, "Information theory and an extension of the maximum likelihood principle", Procs. of the 2nd International Symposium on Information Theory, Akademiai Kiado, Budapest, 1973.

Beginning of Information Criteria to compute the intrinsic dimensionality of a model, i.e. what is the best compromise between Goodness of Fit and parsimony of a model.

• D.G. Kendall. "The diffusion of shape". Advances in Applied Probability, 9(3):428­-430, 1977.

The first definition of shape as understood in computer vision that I could find.

• J. Canny, "A Computational Approach to Edge Detection", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 8, no. 6, pp. 679-698, 1986.

First proper definition of an (intensity) edge detector.

• F.L. Bookstein. "Size and shape spaces for landmark data in two dimensions". Statistical Science, 1(2):181­-242, 1986.
• D.G. Kendall. "Comment to `Size and shape spaces for landmark data in two dimensions'". Statistical Science, 1(2):222­-226, 1986.

Bookstein introduced configurations of landmarks (points with a biological correspondence) from morphometrics to computer vision to explain geometry and deformation. But in fact, Kendall noted that this was an approximation to his own work, more mathematically formal. He reworded his original definition of shape (Kendall 1977) as 'Shape is what remains when location, size, and rotational effects are filtered out'.

• L. Sirovich and M. Kirby. "Low-dimensional procedure for the characterization of human faces". Journal of the Optical Society of America, 4(3):519­-524, Mar 1987.

First time that using PCA in computer vision was proposed (to model images of faces).

• M. Kass, A. Witkin and D. Terzopoulos. "Snakes: active contour models". International Journal of Computer Vision, 1(4):321--331, Nov 1988.

Proposal of active contour models (snakes), and the start of energy minimisation deformable models.

• F.L. Bookstein, "Principal warps: Thin-plate splines and the decomposition of deformations", IEEE Transactions on Pattern Analysis and Machine Intelligence, 11(6), Jun 1989.

The starting point for image non-rigid registration. How to compute dense deformation fields from discrete sets of points.

• J.R. Bergen, P. Anandan, K.J. Hanna and R. Hingorani. "Hierarchical model-based motion estimation", Procs. of Computer Vision — ECCV'92, 588:237-252.

Proposal of using a global model followed by a local model, and using a multiresolution scheme, for image registration (and by extension, segmentation) algorithms. This is the standard heuristic. They argued that it is not only efficient but necessary to ignore high resolution information when computing large displacements.

• T.F. Cootes, C.J. Taylor, D.H. Cooper, and J. Graham. "Training models of shape from sets of examples". In Procs. of the British Machine Vision Conference (BMVC), pp. 266–275, Berlin, 1992. Springer.

Proposed computing a shape space applying PCA to landmark configurations, and called it the Point Distribution Model (PDM)

• W. Wells, P. Viola and R. Kikinis. "Multi-Modal Volume Registration by Maximization of Mutual Information". Proceedings of the 2nd International Symposium on Medical Robotics and Computer Assisted Surgery 1995; 2:55-62.
• W. Wells, P. Viola, H. Atsumi, S. Nakajima and R. Kikinis. "Multi-modal volume registration by maximization of mutual information". Med Image Anal. 1996 Mar;1(1):35-51. PMID: 9873920.

Introduction of Mutual Information (MI) to image registration, that enables registration of images from different modalities (i.e. CT to MRI).