Linear Time Maximally Stable Extremal Regions

Abstract

In this paper we present a new algorithm for computing Maximally Stable Extremal Regions (MSER), as invented by Matas et al. The standard algorithm makes use of a union-find data structure and takes quasi-linear time in the number of pixels. The new algorithm provides exactly identical results in true worst-case linear time. Moreover, the new algorithm uses significantly less memory and has better cache-locality, resulting in faster execution. Our CPU implementation performs twice as fast as a state-of-the-art FPGA implementation based on the standard algorithm.The new algorithm is based on a different computational ordering of the pixels, which is suggested by another immersion analogy than the one corresponding to the standard connected-component algorithm. With the new computational ordering, the pixels considered or visited at any point during computation consist of a single connected component of pixels in the image, resembling a flood-fill that adapts to the grey-level landscape. The computation only needs a priority queue of candidate pixels (the boundary of the single connected component), a single bit image masking visited pixels, and information for as many components as there are grey-levels in the image. This is substantially more compact in practice than the standard algorithm, where a large number of connected components must be considered in parallel. The new algorithm can also generate the component tree of the image in true linear time. The result shows that MSER detection is not tied to the union-find data structure, which may open more possibilities for parallelization.

Automatic Sea Bird Detection From High Resolution Aerial Imagery

“…The detected MSER's are given as pixel arrays masking the actual regions. In our approach we use the OpenCV implementation of the MSER method, which is a variant improved for computational complexity (Nistér, 2008). As such it is able to efficiently handle the image sizes occurring in digital aerial imaging.…”

OpenMVG: Open Multiple View Geometry

“…• Affine invariant regions: Tree-Based Morse Regions (TBMR) [23], Maximally Stable Extremal Regions (MSER) [24].…”

Mobile Visual Location Recognition

“…As the Hessian-Affine detector requires at least four times more computations than the MSER detector, it is reasonable to employ only the latter in mobile application. Optimized implementations like the approach proposed in [NS08] allow us to process an image of 640x480 pixels in 50 ms on a 2 GHz CPU, which can be further improved with the aid of parallel processing approaches on the GPU of the mobile device. Tracking these regions and computing additional MSERs only from the previously unseen boundary textures allows for a 5-6 fold speedup [TCGP09,DB06].…”

“…Finally, the proposed algorithm is simple and efficient. MSER as well as the distance transform can be very efficiently computed [NS08,Bai04] and determining the stroke width requires only a lookup table (Section 6.1.3).…”

“…At the input of the system, the image intensities are linearly adjusted to enhance the contrast. Subsequently, MSER regions are efficiently extracted from the image [NS08] and enhanced using Canny edges obtained from the original gray-scale image (Section 6.1.1). As a next step, the resulting CCs are filtered using geometric constraints on properties like aspect ratio and number of holes (Section 6.1.2).…”