In this paper, we propose a new approach in image processing for interest point detection. Thanks to the link between linear scale-space and wavelet decompositions, we are able to build a structure in scale-space based on wavelet maxima lines. This framework allows robust detection of objects (or blob-like structures) present in the image, and the computation of their characteristic scale. Eventually the estimation of their shape can be performed. Keywords: scale-space representation, maxima lines of wavelet decomposition, blob detection, scale selection.