Dwarf galaxies are the most common galaxies in the Universe. They are systems believed to consist of matter in a near-primordial state, from which giant galaxies probably form. As such, they are important probes for studying matter in its near-primordial state. In an effort to study the main physical and chemical properties of dwarfs, the present thesis focuses upon the main physical properties of dwarfs. Two classes of star forming dwarf galaxies are considered: dwarf irregulars (dIs), and blue compact dwarfs (BCDs). A third class, dwarf ellipticals (dEs), is studied based on its structural properties and compared with dIs. Possible evolutionary connections are addressed between dIs and BCDs. To measure the luminosity, deep imaging in the near-infrared (NIR) is considered. Compared with the visible, the NIR domain gives a better gauge of the galaxy mass contained in the old stellar populations, minimising the starburst contribution and also the effects of extinction. Two observing samples of star-forming dwarf galaxies are considered. The first includes 34 dIs in the Local Volume closer than 5 Mpc. The second sample includes 16 BCDs in the Virgo Cluster. In six observing runs between 2001 and 2004, we acquired deep NIR images (J and K_s) using the 3.6m Canada-France-Hawaii-Telescope (CFHT) in Hawaii and the 2.1m telescope at the National Astronomical Observatory ''San Pedro Martir'' (OAN-SPM) in Mexico. Deep spectrocopy was acquired in 2003 on the 8.1m Gemini-North telescope in Hawaii. We completed the observed samples with spectroscopic data from the literature, and photometry from the 2MASS survey and GOLDMine database. From a statistical study at CFHT, we derived some strategies necessary to image optimally faint extended sources in the NIR. Due to the airglow variation in the atmosphere and the thermal contribution of the dome, telescope and the instrumentation, repeated observations of the sky must be alternated every 3-4 minutes with the science images, in order to achieve 1% accuracy in surface photometry. We discovered that the NIR surface brightness profiles of dIs can be fitted with a hyperbolic secant (sech) function with only two parameters: the central surface brightness and the scale length. This led to the discovery that BCD profiles could be fitted with a sech component to trace the diffuse component responsible for most of the light, and a Gaussian for the central starburst. For 25 of the 34 observed dIs, we resolved stars as faint as M_K=-7.5 mag out to 5 Mpc. We show that the resolved component comprises more than 50% of the light from star formation bursts within the last 3 Gyr. By separating the resolved sources associated with each galaxy from the unresolved component, we determined for the first time the contribution from the resolved stellar component to the total light in the NIR. In nearly all galaxies, the resolved population up to M_K=-7.5 mag represents less than 5% of the total flux in K_s, with ratios in J 1.5-2 times larger. Compared with the visible, the small contribution of the resolved flux allows us to consider the NIR a better domain to sample the old stellar populations, and thus use it to gauge the stellar mass of star-forming dwarf galaxies. For 29 dIs, colour-magnitude diagrams (CMDs) for the resolved component were derived. Three CMDs include more than 1000 stars in both K_s and J, while another 15 CMDs have more than 100 stars. Most of the CMDs show a main blue finger centered around J-K_s=1 mag. In some cases, a red tail extends from the finger out to J-K_s =+2.5 mag. The colour profiles of the unresolved components show a remarkably constant J-K_s = +0.8 to +1.0 mag, which matches the colour of the main finger in the CMDs. For both dIs and BCDs, we searched for correlations between galaxy size, absolute magnitude, central surface brightness, colours, and the resolved over total ratio (for dIs). Good linear correlations were found between the scale length, and the sech magnitude, and between the isophotal semimajor axis and the sech magnitude. Also, correlations were found between the central surface brightness and the sech magnitude. Overall, galaxies with more luminous old components are larger, redder, and brighter in the centre. Thus, size, colour, and the extent of the central plateau appear to be determined by the mass of the old component. For both dIs and BCDs, the Tully-Fisher relation shows considerable scatter in K_s, especially at low luminosities. The scatter appears to be tied to variations in surface brightness. A new ''fundamental plane'' was discovered for dIs which relates the sech absolute magnitude, the central surface brightness, and the neutral hydrogen line-width. The residuals are low enough (rms 0.4 mag), that it offers considerable potential as a distance indicator for star-forming dwarfs. BCDs appear to lie on the dI fundamental plane, but the scatter is larger, probably due to uncertitudes in their line widths. We used NIR images for 22 dEs in the Virgo cluster taken from the GOLDMine database, and some additional data for 9 other dwarf spheroidals in the Local Group, to examine how closely dEs fit into the dI fundamental plane. Over a 9 mag interval in absolute magnitude, the dEs fall in the plane defined by the dIs. The outstanding overlap suggests a close evolutionary connection between dIs and dEs. Using oxygen abundances of dIs and BCDs from the literature, we studied correlations between metallicity, stellar mass, gas mass, baryonic mass, and gas fractions. Although there is some scatter, metallicity correlates with all four parameters in the sense that more massive systems contain more metals. The oxygen abundance correlates very well with the luminosity in K_s, for both dIs and BCDs. Nevertheless, the two relations appear to be different, with more luminous BCDs being more metal rich than dIs with similar absolute magnitudes. This shift can be biased by not including gas. After the gas mass is considered, BCDs allign with dIs on the same mass-metallicity relation, suggesting similar evolutionary connections. Overall, based on their structural and physical properties, closer links emerge between dIs, BCDs, and dEs. Comparing their structural properties, dIs and BCDs appear to share common origins, with the sech component modeling all the NIR flux for dIs and most of the flux for BCDs. On the fundamental plane, dIs and BCDs also show similar dynamical properties. Thus, BCDs seem to be dIs observed in a bursting phase. Taking into account their gas-to-mass fraction, BCDs cannot be considered closed systems, their shifted position in respect to dIs suggesting flows of gas such as infall of gas clouds. Comparing structural properties, dEs lie in the fundamental plane of dIs, suggesting an intimate link between the two systems, in the sense that dEs could be consider the final outcome of dIs after all gas is removed from the system. Full thesis available online at http://aries.phys.yorku.ca/ ovidiuv/Thesis.pdf (7 MB, 247 pag)