Tea [Camellia sinensis (L.) O. Kuntze] has long history in production and consumption in Zhejiang Province, China. Improvement of tea, therefore, is of great importance and a good understanding of the genetic diversity and population structure of tea germplasm is a prerequisite to the improvement. In spite of great advances on the use of molecular markers in tea, achievement is still gotton very slowly compared with in other cereal crops and woody species. Expressed sequence tag derived simple sequence repeat (EST-SSR) is a less costly alternative way of developing new markers for genetic diversity analysis, functional markers development and marker assisted breeding of tea plant. A total of 4 833 ESTs generated from a cDNA library of tea young root were subjected to SSR mining using DNAstar 5.0 software, 577 EST-SSRs were identified and 416 primer pairs were designed by Primer premier 5.0. After the determination of annealing temperatures and polymorphism of all the primers, 64 core primers were selected and used for genetic diversity and population structure analyses of tea landraces and improved cultivars in Zhejiang province. All selected primers were polymorphic and 232 alleles were amplified, with 3.6 alleles per primer pair on an average. Each primer pair identified 2 to 13 genotypes, with an average of 4.3. The mean of polymorphism information content (PIC) was 0.44, ranging from 0.02 to 0.84. Observed heterozygosity (H_o) was 0.44, while expected heterozygosity (H_e) was 0.48. The level of genetic diversity among landraces was slightly higher than that among improved cultivars and breedinglines. There were 226 alleles amplified in 22 landraces with 14 of them that were special. In the thirty-seven improved cultivars, however, two hundred and eighteen alleles were amplified but only six were special. The PIC of the landrace groups varied from 0.24 to 0.36, in which Juyan Qunti was the highest and Huiming Qunti was the lowest. Deqing Qunti was closest to Juyan Qunti in genetic relationship, but farthest from Huiming Qunti. The genetic diversity of tea cultivars from Hangzhou was the highest with PIC of 0.41, while those from Lishui recorded the lowest PICof 0.24. Population structure revealed by software Structure 2.2 and UPGMA cluster analysis showed that landraces and improved cultivars were relatively independent. The improved cultivars were further clustered into smaller groups according to their pedigree. Hybrid offspring from Fuding Dabaicha and Yunnan Dayezhong from different breeding organizations fell into similar group.