Search

Because noise can interfere with downstream analysis, image denoising has come to occupy an important place in the image processing toolbox. The most accurate state-of-the-art denoisers typically train on a representative dataset. But gathering a training set is not always feasible, so interest has grown in blind zero-shot denoisers that train only on the image they are denoising. The most accurate blind-zero shot methods are blind-spot networks, which mask pixels and attempt to infer them from their surroundings. Other methods exist where all neurons participate in forward inference, however they are not as accurate and are susceptible to overfitting. Here we present a hybrid approach. We first introduce a semi blind-spot network where the network can see only a small percentage of inputs during gradient update. We then resolve overfitting by introducing a validation scheme where we split pixels into two groups and fill in pixel gaps using domino tilings. Our method achieves an average PSNR increase of $0.28$ and a three fold increase in speed over the current gold standard blind zero-shot denoiser Self2Self on synthetic Gaussian noise. We demonstrate the broader applicability of Pixel Domino Tiling by inserting it into a preciously published method.

We present an expanded kinematic study of the young cluster NGC 2264 based upon optical radial velocities measured using multi-fiber echelle spectroscopy at the 6.5 meter MMT and Magellan telescopes. We report radial velocities for 695 stars, of which approximately 407 stars are confirmed or very likely members. Our results more than double the number of members with radial velocities from F{\H u}r{\'e}sz et al., resulting in a much better defined kinematic relationship between the stellar population and the associated molecular gas. In particular, we find that there is a significant subset of stars that are systematically blueshifted with respect to the molecular ($^{13}$CO) gas. The detection of Lithium absorption and/or infrared excesses in this blue-shifted population suggests that at least some of these stars are cluster members; we suggest some speculative scenarios to explain their kinematics. Our results also more clearly define the redshifted population of stars in the northern end of the cluster; we suggest that the stellar and gas kinematics of this region are the result of a bubble driven by the wind from O7 star S Mon. Our results emphasize the complexity of the spatial and kinematic structure of NGC 2264, important for eventually building up a comprehensive picture of cluster formation., Comment: Accepted to AJ. 38 pages, 5 Figures 3 Tables

We extend our previous study of the stellar population of L1641, the lower-density star-forming region of the Orion A cloud south of the dense Orion Nebula Cluster (ONC), with the goal of testing whether there is a statistically significant deficiency of high-mass stars in low-density regions. Previously, we compared the observed ratio of low-mass stars to high-mass stars with theoretical models of the stellar initial mass function (IMF) to infer a deficiency of the highest-mass stars in L1641. We expand our population study to identify the intermediate mass (late B to G) L1641 members in an attempt to make a more direct comparison with the mass function of the nearby ONC. The spectral type distribution and the K-band luminosity function of L1641 are similar to those of the ONC (Hillenbrand 1997; Muench et al. 2002), but problems of incompleteness and contamination prevent us from making a detailed test for differences. We limit our analysis to statistical tests of the ratio of high-mass to low-mass stars, which indicate a probability of only 3% that the ONC and the southern region of L1641 were drawn from the same population, supporting the hypothesis that the upper mass end of the IMF is dependent on environmental density., Comment: 14 pages, 11 figures, accepted for publication in ApJ

We present results from an optical photometric and spectroscopic survey of the young stellar population in L1641, the low-density star-forming region of the Orion A cloud south of the Orion Nebula Cluster (ONC). Our goal is to determine whether L1641 has a large enough low-mass population to make the known lack of high-mass stars a statistically-significant demonstration of environmental dependence of the upper mass stellar initial mass function (IMF). Our spectroscopic sample consists of IR-excess objects selected from the Spitzer/IRAC survey and non-excess objects selected from optical photometry. We have spectral confirmation of 864 members, with another 98 probable members; of the confirmed members, 406 have infrared excesses and 458 do not. Assuming the same ratio of stars with and without IR excesses in the highly-extincted regions, L1641 may contain as many as ~1600 stars down to ~0.1 solar mass, comparable within a factor of two to the the ONC. Compared to the standard models of the IMF, L1641 is deficient in O and early B stars to a 3-4 sigma significance level, assuming that we know of all the massive stars in L1641. With a forthcoming survey of the intermediate-mass stars, we will be in a better position to make a direct comparison with the neighboring, dense ONC, which should yield a stronger test of the dependence of the high-mass end of the stellar initial mass function upon environment., Comment: 19 pages, 21 figures. Accepted by ApJ

We report a set of numerical experiments aimed at addressing the applicability of competitive accretion to explain the high-mass end of the stellar initial mass function in a sheet geometry with shallow gravitational potential, in contrast to most previous simulations which have assumed formation in a cluster gravitational potential. Our flat cloud geometry is motivated by models of molecular cloud formation due to large-scale flows in the interstellar medium. The experiments consisted of SPH simulations of gas accretion onto sink particles formed rapidly from Jeans-unstable dense clumps placed randomly in the finite sheet. These simplifications allow us to study accretion with a minimum of free parameters, and to develop better statistics on the resulting mass spectra. We considered both clumps of equal mass and gaussian distributions of masses, and either uniform or spatially-varying gas densities. In all cases, the sink mass function develops a power law tail at high masses, with $dN/dlog M \propto M^{-\Gamma}$. The accretion rates of individual sinks follow $\dot{M} \propto M^2$ at high masses; this results in a continual flattening of the slope of the mass function towards an asymptotic form $\Gamma \sim 1$ (where the Salpeter slope is $\Gamma = 1.35$). The asymptotic limit is most rapidly reached when starting from a relatively broad distribution of initial sink masses. In general the resulting upper mass slope is correlated with the maximum sink mass; higher sink masses are found in simulations with flatter upper mass slopes. Although these simulations are of a highly idealized situation, the results suggest that competitive accretion may be relevant in a wider variety of environments than previously considered, and in particular that the upper mass distribution may generally evolve towards a limiting value of $\Gamma \sim 1$., Comment: 20 pages, 12 figures