I don't think you can completely get rid of histogram or threshold-based binarization, since the former is to achieve line segmentation, while the latter is to extract the letters. The Radon horizontal projection is used for line segmentation, and the center line can be used to approximate the baseline of each segment. Yet this is somehow equivalent to the horizontal histogram.
There are some improvement based on the basic histogram. In Zheng et al's paper A model based line detection in documents, they applied a directional single connected chain strategy that considers the global textline orientation and merges the non-overlapped small segments into one line. The algorithm provides more accurate base line detection than the pure histogram-based line segmentation.
Besides, in Arabic script where short characters and large diacritics exist, the Tarik et al's paper A Novel Baseline Detection Method of Handwritten Arabic-Script Documents Based on Sub-Words that measures the vertical distance between the semi-overlapped subwords baselines.
Yet in those studies in the attempt to increase the accuracy, histogram and/or threshold were still a necessity.